Abstract: Flow identification plays an important role in guaranteeing the performance of quality of service for digital home networks. On the basis of characteristic analysis and flow level five tuple analysis, this paper proposes Flow Level characteristic approaches and studies the application features of Video On Demand（VOD）, QQ video, download and multi-screens etc. from the perspective of five tuple, the ratio of total number of uplink packets to total number of downlink packets, uplink and downlink mean packet length and its ratio, the ratio of uplink total bytes to downlink total bytes, the ratio of the number of source addresses to the number of source ports. Furthermore the P2P VOD and P2P download identification are further studied. This Flow Level characteristic analysis approaches are satisfied with the need of flow identification for digital home network. In the experiment, the bandwidth occupation is measured, the features of the main applications per unit time are analyzed and extracted in the home networks. The multi-application real-time identification system for the multi-devices on the Linux platform is realized.

Key words: digital home networks, application identification, flow level characteristic approaches, real-time flow identification system

摘要： 数字家庭网络流量识别对保障数字家庭网络服务质量（Quality of Service，QoS）具有重要作用。提出的Flow Level 流量特征分析法结合Flow Level五元组分析和特征值分析，从五元组，上下行包数比，上下行平均包长及比值，上下行数据量比，源端地址数和端口数之比等方面研究视频点播、视频通话、下载和无线传屏等应用特征。其中，对P2P（Peer to Peer）视频点播和下载的识别方案做了进一步研究。最终满足了数字家庭网络流量识别的需求。在仿真实验阶段，测得家庭网络环境中主要应用占据带宽情况，分析并提取单位时间内应用流量特征值。最终基于Linux平台实现了多终端设备的多应用实时识别系统。

关键词: 数字家庭网络, 应用识别, Flow Level流量特征分析法, 流量实时识别系统